Atomic modeling of the ITS2 ribosome assembly subcomplex from cryo‐EM together with mass spectrometry‐identified protein–protein crosslinks

Abstract

The assembly of ribosomal subunits starts in the nucleus, initiated by co‐transcriptional folding of nascent ribosomal RNA (rRNA) transcripts and binding of ribosomal proteins and assembly factors. The internal transcribed spacer 2 (ITS2) is a precursor sequence to be processed from the intermediate 27S rRNA in the nucleoplasm; its removal is required for nuclear export of pre‐60S particles. The proper processing of the ITS2 depends on multiple associated assembly factors and RNases. However, none of the structures of the known ITS2‐binding factors is available. Here, we describe the modeling of the ITS2 subcomplex, including five assembly factors Cic1, Nop7, Nop15, Nop53, and Rlp7, using a combination of cryo‐electron microscopy and cross‐linking of proteins coupled with mass spectrometry approaches. The resulting atomic models provide structural insights into their function in ribosome assembly, and establish a framework for further dissection of their molecular roles in ITS2 processing.

Introduction

Ribosome biogenesis in yeast starts with the transcription of the ribosomal DNA into a single 35S primary transcript (∼6600 nucleotides), including sequences for both the large (5.8S and 25S) and small subunits (18S), as well as internal and external spacer sequences to be processed along the maturation pathways.1 The internal transcribed spacer 2 (ITS2) is a stretch of precursor sequence that separates the 5.8S and 25S rRNA. The processing of ITS2 is initiated by an endonucleolytic step, followed by further trimming through collaborative effects of multiple RNases in two directions to form the mature 3′‐ and 5′‐ends of the 5.8S and 25S rRNAs, respectively.2 The proper processing of ITS2 is also dependent on multiple ITS2‐binding factors and is coupled with the release of early assembly factors (such as a few A3 and B factors) that are distant from the ITS2 sequence.1 Known non‐ribonuclease factors that directly associate with ITS2 include Nop15, Rlp7, Cic1,3 ⁻⁵ and an RNA helicase Mtr4.6 In addition, Erb1, Nop7, and Nop53 were reported to directly bind to 25S rRNA helices in the neighborhood of ITS2.3, 6 The processing of ITS2 is not merely a removal of spacer sequences, but more importantly it is coupled with quality control pathways, as mutations of certain ITS2‐binding A3 factors lead to rapid turnover of pre‐rRNAs.1, 2 Nevertheless, the structures of ITS2 factors and the underlying mechanisms by which they function remain largely unknown.

Ribosome biogenesis is an extraordinarily complex process involving over 200 trans‐acting protein factors, which are organized into distinct temporal and spatial groups to facilitate and to regulate the maturation of ribosomal subunits. Cryo‐electron microscopy (cryo‐EM) is an ideal tool for structural characterization of purified endogenous pre‐ribosomes7, 8, 9, 10 or in vitro reconstituted factor‐containing ribosomal complexes,11, 12, 13, 14, 15 as they are usually highly heterogeneous in both composition and structure. With the recent development of detector technology and powerful algorithms for image processing, the cryo‐EM single‐particle approach is now entering a stage with a fast growing number of atomic or near‐atomic structures being produced.16, 17, 18

The atomic modeling of cryo‐EM maps has been very challenging, considering the non‐uniform resolution of the maps, particularly when the large complexes contain many flexible and transiently associated components. Chemical cross‐linking of proteins coupled with mass spectrometry (CXMS) has become a robust tool for analyzing structures of multiple proteins in various mixtures.19 In the last decade, CXMS gained its popularity in structural studies because compared with traditional approaches it requires less on sample, is more tolerant to sample purity, and is good at providing structural information of highly dynamic regions. It usually involves a residue‐specific chemical cross‐linker to form covalent bonds between two amino acid residues in close proximity, and then mass spectrometry to analyze cross‐linked peptides and locate cross‐linking sites with dedicated identification software. CXMS‐derived information can nicely complement traditional structural analysis in deciphering the architecture of protein complexes. In particular, the last decade has witnessed the successful marriage of cryo‐electron microscopy and CXMS; the former generates the molecular shape of protein complexes and the latter provides distance constraints at the residue level. A growing number of macro‐molecular architectures have been studied using this hybrid approach, among them the 26S proteasome,20 the nuclear pore complex,21 the RNA polymerase II transcription pre‐initiation complex,22 the INO80 chromatin remodeler complex,23 and the eukaryotic translation initiation complex24 (for more examples, see Ref. 25).

Recently, we have reported an atomic structure of nuclear pre‐60S ribosomal particles from Saccharomyces cerevisiae, purified through an epitope‐tagged assembly factor Nog2.26 Here, we describe the modeling process of the ITS2 region of this pre‐60S cryo‐EM map (3.08 Å), where the structural and compositional information are nearly completely unknown. With a combination of multiple approaches, including cryo‐EM, CXMS, secondary, and tertiary structural prediction, we have built partial atomic models for five assembly factors (Cic1, Nop7, Nop15, Nop53, and Rlp7). The structures and their implications in ITS2 folding and processing are subsequently discussed.

Results

Overview of the cryo‐EM map of the ITS2 region

The pre‐60S particles were purified through an epitope‐tagged Nog2,26 an essential assembly GTPase that functions primarily in the nucleoplasm.27 Cryo‐EM single particle analysis of the Nog2‐particles rendered a density map at the nominal resolution of 3.08 Å.26 Similar to the previous low‐resolution cryo‐EM structure of pre‐60S particles isolated with tagged Arx1,8, 9 the structure of the Nog2‐particles features a “foot‐like” region (Fig. 1), presumably being the location of the sequences of ITS2 and associated protein factors.

Figure 1.

Structural overview of the pre‐60S particle and its “foot‐like” region. (A): Cryo‐EM map of the pre‐60S particle with unassigned density colored dark gray and density corresponding to the mature 60S subunit colored light gray. (B): Previously known assembly factors are highlighted in different colors. (C): The segmented map of the “foot‐like” region of the pre‐60S particle. (D): Two views of the “foot‐like” region are shown, with density of the ITS2 rRNA colored yellow.

Although the locations of a few assembly factors on pre‐60S ribosomes, such as Rsa4, Mrt4, Nog1, Tif6, Rlp24, and Arx1, have been reported in previous cryo‐EM and crystallography studies [Fig. 1(B)],8, 10, 12, 14, 28, 29 the structure and the exact composition of the ITS2 region are completely unknown. The quality of the density map allows us to separate the rRNA from proteins with high confidence [Fig. 1(D)], as some of the ITS2 nucleotides form a helix which displays a typical ribbon‐like shape. Although the density for proteins is also nicely resolved, two problems were encountered for atomic modeling of the ITS2 subcomplex. One is that over 30 assembly factors were reported to exist in the Nog2‐particles based on the mass spectrometry data,26 and no high‐resolution structure exists for any of the known ITS2‐associated factors. The other is that the factors in the ITS2 region might not be resolved for their full‐length. The intertwined polypeptide chains and often disordered loops make the backbone tracing rather challenging. To overcome these obstacles, we set out to seek help from CXMS, secondary and tertiary structural prediction. Structural predictions were therefore performed for all the factors identified in the mass spectrometry data, using PSIPRED30 and I‐TASSER31 for the 2D and 3D predictions, respectively.

Connectivity map of ITS2 subcomplex by CXMS

To obtain connectivity information between assembly factors and the 60S ribosomal core, we performed CXMS experiments on the Nog2‐particles, where 219 proteins were identified at 0.48% protein false discovery rate (FDR). Two lysine‐targeted cross‐linkers—DSS and its water‐soluble analog BS³—were used in two independent cross‐linking reactions. All the MS/MS spectra were searched against a database consisting of the 219 proteins. By requiring FDR <5%, E‐value <0.00001, and spectral count ≥2, a total of 282 non‐redundant cross‐linked lysine pairs were identified, comprising 182 intra‐protein and 100 inter‐protein cross‐links. To evaluate the quality of the cross‐links, we mapped them to the crystal structure of the 60S ribosomal subunit (PDB code: 3u5e).32 DSS and BS3 both provide a distance constraint of 24 Å, which is the sum of the 11.4‐Å spacer arm and the flexible 6‐Å side chains of the two lysine residues. Out of 105 cross‐links whose connected residues are both present in the crystal structure, 98 have Cα–Cα distances ≤24 Å. This result suggests that the cross‐links we identified reflect spatial information reliably. Cross‐links with distances of more than 24 Å could either be derived from conformational flexibility or due to false positive identification. Of particular interest to us are the linkages bridging a ribosomal protein and an assembly factor, as they are valuable for placement of the assembly factors relative to the 60S ribosomal core (Supporting Information Table S1). In this work, we focus on the location of five assembly factors, Cic1, Nop7, Nop15, Rlp7, and Nop53, because they form a cluster of cross‐links at the expected region of the ITS2 (Fig. 2).

Figure 2.

Cross‐linking data related to the “foot‐like” region. (A): Three assembly factors (Cic1, Nop15, and Nop53) are cross‐linked to ribosomal proteins around the “foot‐like” region. Red dots represent cross‐linking sites in ribosomal proteins. For clarification, only structures of the ribosomal proteins (gray) from the crystal structure of the yeast ribosome (PDB id: 3u5e)32 are shown in ribbon representation, with those involved in cross‐linking to assembly factors highlighted in wheat. L25(K6), L8(K21), and L22(A2) are absent in the structure, so their closest residues in sequence are shown instead. (B): Eleven protein–protein cross‐links are shown involving five assembly factors and six ribosomal proteins.

These five proteins were found to have 8 intra‐protein and 15 inter‐protein cross‐links, which are translated into 11 protein–protein interactions [Fig. 2(B), Supporting Information Table S1]. Eight cross‐links indicate direct contact between an assembly factor (Cic1, Nop15, or Nop53) and the ribosomal proteins [Fig. 2(A)]. Of these five factors, Nop53 appears to be central, as it displays the most cross‐links with both factors and ribosomal proteins [Fig. 2(B)]. This is highly consistent with the fact that Nop53 is an important adaptor protein responsible for recruiting the exosome machinery6, 33 to process the 7S pre‐rRNA, through interaction with the Exosome‐associated helicase Mtr4.6 The detection of Nop7 in the ITS2 cluster (cross‐linking to Nop53) also agrees with previous data that Nop7 assembles into 90S particles containing 35S pre‐rRNA and, like Cic1, Rlp7, and Nop15, is necessary for conversion of 27SA₃ pre‐rRNA to 27SB_S pre‐rRNA.34

Positioning of core domains of Cic1, Nop15, and Rlp7 in the ITS2 region

The CXMS data show that Cic1 and Nop15 are cross‐linked to L8 (K96) and L25 (K6), respectively [Fig. 2(A) and Supporting Information Table S1]. Based on the crystal structure of the yeast ribosome,32 these two positions are close to the ITS2 rRNA density identified in the cryo‐EM map [Fig. 2(A)], suggesting that they likely directly bind to the ITS2. Therefore, we decided to locate these two factors first in the map regions close to the ITS2 rRNA. Combining both the sequence and structural prediction analyses of Cic1 and Nop15, we found that Nop15 contains an RNA recognition motif (RRM) [Fig. 3(A)] and Cic1 belongs to the ribosomal protein L1‐like family which consists of one alpha/beta subdomain interrupted by another alpha/beta subdomain [Fig. 3(C)]. Therefore, the predicted RRM domain of Nop15 was cut out from the predicted 3D structure, and manually fitted into the density map based on the overall shape match [Fig. 3(A)]. For Cic1, in a manner similar to that used for Nop15, two subdomains from the predicted 3D structure were selected and separately fitted as rigid bodies [Fig. 3(C)]. As expected, the positions of the three subdomains were located, as they display relatively high correlation coefficients (above 0.8). Subsequently, the fitted structures of Nop15 and Cic1 were adjusted in Coot,35 using the bulky residues (such as Trp, Arg, Phe, and Tyr) as markers for primary sequence assignment, followed by backbone tracing in both the N‐ and C‐terminal directions. As a result, we were able to build a model for a majority of sequences for Nop15 (residues 88–220, full‐length 220 residues) and Cic1 (residues 31–51 and 71–305, full‐length 376 residues) [Fig. 3(B,D)].

Figure 3.

Modeling of Cic1 and Nop15. (A): The predicted model of the RNA recognition motif (residues 91–169) of Nop15 is fitted into the map as a rigid body. (B): Final atomic model of Nop15 (residues 88–220) superimposed with the map. (C): The predicted models of two α/β subdomains of Cic1 are fitted into the density map as two rigid‐bodies. (D): Final atomic model of Cic1 superimposed with the map. (E): The C‐terminal helix of Nop15 interacts with Cic1. (F): The spatial relationship between Nop15 and Cic1 in the ITS2 region. Density of the ITS2 RNA is colored yellow.

Next, given the CXMS data that K54 and K64 of Rlp7 are cross‐linked to K109 of Nop15, we set out to find identifiable structural features in the proximity of Nop15‐K109 in the cryo‐EM density map. Very interestingly, the secondary structural prediction of the N‐terminal sequence of Rlp7 shows α‐helices exclusively, with a high tendency to form a very long helix [Fig. 4(A)]. Consistently, the cryo‐EM map near the region of Nop15‐K109 also features a long bent α‐helix, which allowed us to accurately assign the residues 15–88 of Rlp7 in the density map [Fig. 4(A)]. Structural prediction of Rlp7 indicates that it belongs to the ribosomal protein L30p/L7e family and contains a ferredoxin‐like‐fold domain. Similarly, we located the position of this subdomain of Rlp7 with rigid‐body fitting in the density map [Fig. 4(B)]. Further optimization of the model by Coot rendered a partial atomic model for Rlp7 (residues 15–105 and 127–322, full‐length 322 residues) [Fig. 4(C)]. In good agreement with previous cross‐linking data,4, 5 Rlp7 directly binds to the junction of the 5′‐end of the 25S, ITS2, and the 3′‐end of the 5.8S rRNAs [Fig. 4(D)].

Figure 4.

Modeling of Rlp7. (A): A long helix in the N‐terminal region of Rlp7 is identified in the map according to its cross‐linking sites with Nop15. The secondary structure prediction of Rlp7 is shown in the lower right panel. (B): The predicted ferredoxin‐like fold domain of Rlp7 is fitted into the density map as a rigid body. (C): Final atomic model of Rlp7 superimposed with the map. (D): Rlp7 directly interacts with the linkage region of the 5.8S, ITS2, and 25S rRNAs (marked by asterisk). The ITS2 rRNA is colored yellow.

Ab initio modeling of Nop53 guided by CXMS

For Nop53, unfortunately, the 3D structural prediction produced no reliable, compact subdomains. Therefore, we performed ab initio modeling of Nop53 using the distance information from the CXMS data as a constraint for secondary structure searching in the density map. K257 and K259 of Nop53 are cross‐linked to K116 of L27 and K21 of L8, respectively (Fig. 5). Both K257 and K259 are located in a predicted long α‐helix formed by residues 231–266. Indeed, within 12 Å from K116 of L27, an unassigned density rod was found, which could perfectly account for the predicted α‐helix of residues 231–266 (Fig. 5, right panel). Another piece of Nop53 was located by the data that Nop53‐K192 could be cross‐linked to K14 of L35. Nop53‐K192 is within a predicted short helix (residues 187–192), which consequently led to the localization of this short helix in the density map (Fig. 5, left panel). This assignment of helix 187–192 is further confirmed by the bulky side chain of downstream W197. With these two α‐helices fixed and assigned in the density map, we could extend the chain of Nop53. At last, we were able to build models for two pieces of Nop53 (residues 167–272, and 379–455, full‐length 455 residues).

Figure 5.

Modeling of Nop53. Two helices identified in the density map according to the cross‐linking of Nop53 to ribosomal proteins L35 and L27, respectively. The sites of cross‐linking and distances are labeled.

Modeling of Nop7 in the remaining density of the ITS2 region

After finishing modeling of the above‐mentioned four factors, a significant portion of the map is still left unassigned. A poly‐alanine model was then built. As Nop7‐K79 is cross‐linked to Nop53‐K170, we started to compare the pattern of predicted secondary structures of the N‐terminal sequences of Nop7 with the poly‐alanine model. Because the side chains are nicely resolved in this region and a good match between the model and the secondary features of the map was found, we were able to assign residues 1–267 of Nop7 into the density map [Fig. 6(A,B)]. Unfortunately, no obvious density can be traced beyond residue 267 of Nop7. Nevertheless, the predicted BRCT domain of the C‐terminal sequences of Nop7 was found to have a good fit to the remaining density [Fig. 6(C)]. Consequently, models of two parts of Nop7 (residues 1–267, 351–396, and 403–460, full‐length 605) were built after the main chain extension [Fig. 6(D,E)].

Figure 6.

Modeling of Nop7. (A): Cryo‐EM density of a representative region of Nop7, superimposed with the atomic model. The predicted secondary structure of the N‐terminal region of Nop7 is shown on the right panel. (B): The poly‐alanine skeleton of the N‐terminal half of Nop7. (C): The predicted BRCT domain of Nop7 is fitted into the density map as a rigid body. (D): Final model for the BRCT domain of Nop7. (E): The final atomic model of Nop7 after chain extension.

Functional implication of the ITS2‐associated factors in rRNA processing

Based on their spatial relationship with the ITS2 rRNA, these five factors can be categorized into two groups. The first three, Cic1, Nop15, and Rlp7, all directly interact with the ITS2 rRNA through their RNA binding motifs/domains (Figs. 3 and 4). Therefore, they may function to facilitate local folding of the ITS2 nucleotides. Nop15 and Cic1 bind to the two different sides of the structured ITS2 (Fig. 3F), but Nop15 contains a long C‐terminal extension that makes a 90° turn near P192, resulting in interaction with Cic1 through its C‐terminal helix (Fig. 3E). Notably, although Cic1 has been previously shown to be an adaptor protein for the proteasome machinery,36 its association with Nop15 has been reported in previous large‐scale proteomic analysis.37

Given the tight binding of these three factors to the ITS2 rRNA, it can be reasoned that the progressive processing of the ITS2 sequence would need them to dissociate. The departure of these factors could be spontaneous on the basis of local conformational changes on the structured ITS2 rRNA, presumably caused by stepwise removal of nucleotides from the 3′‐end of the 7S rRNA (5.8S plus part of the ITS2 sequence) by the exosome and other nucleases.2 Especially, the N‐terminal helix–turn–helix motif of Rlp7 harbors the binding sites for both the 5′‐ITS2‐25S‐3′ and the 5′‐5.8S‐ITS2‐3′ boundary (Fig. 4D), suggesting that it might be the last remaining factor during the course of ITS2 processing.

In contrast, the other two factors, Nop53 and Nop7, do not interact with the ITS2 rRNA (at least in the ITS2 structure we have built here). But they do have multiple interactions with the mature part of the pre‐60S particle, including both the 25S rRNA and 60S ribosomal proteins. Another interesting observation is that both of them contain long terminal extensions, inserted into the mature part of the pre‐60S particle, which suggests that these terminal extensions might be responsible for anchoring them to the pre‐60S particle. The N‐terminal 120 residues of Nop7 (all well resolved) fold into an exclusively α‐helical structure, and form a very large interface with several helices from the 25S and 5.8S rRNAs (Fig. 7). The extreme N terminus of Nop7 is inserted into a pocket formed exclusively by rRNA. Previous data have shown that the N‐terminal end of Nop7 is important for its function in ribosome biogenesis.38 As to Nop53, its C‐terminal residues 426–455 are embedded in the space on the surface of the mature part of the pre‐60S particle, with its very C‐terminal helix interacting with L25 [Fig. 8(B)]. These structural observations suggest that Nop53 and Nop7 are adaptor proteins, likely responsible for forming a scaffold to recruit additional assembly factors and enzymes required for the ITS2 processing.

Figure 7.

The N‐terminal helical region of Nop7 interacts with multiple rRNA helices. M1 represents the first amino acid (methionine) of Nop7. Different rRNA helices that interact with Nop7 are highlighted and labeled in separate colors.

Figure 8.

Interaction of Nop53 with the rRNA and ribosomal proteins. (A): Nop53 interacts with two assembly factors (Nop7 and Rlp7). (B): Nop53 interacts with three ribosomal proteins (L35, L25, and L27). L25‐N denotes the N terminus of L25. (A) and (B) are rotated 180° about the vertical axis. (C): Both the N‐ and C‐termini of Nop53 interact with multiple rRNA helices. These rRNA helices involved in interactions with Nop53 are highlighted and labeled in separate colors.

The adaptor function of Nop53 has recently been experimentally demonstrated.6 It was shown that the N‐terminal sequence of Nop53, which harbors a common motif with Utp18, is responsible for interacting with Mtr4 to recruit the exosome to the ITS2 region. Although the N‐terminal sequences of Nop53 were not resolved in our map, the structure of Nop53 itself manifests a possible role in scaffolding. It displays a tetrahedral shape with four vertices [Fig. 8(A)]. These four parts interact with totally different components of the pre‐60S particle [Fig. 8(B,C)]. Given that we only built around 40% of its sequence, Nop53 might be able to make contact to an even larger number of factors.

Nop7, together with Ytm1 and Erb1, forms a subcomplex required for 27S rRNA processing.34 The BRCT domain of Nop7 or its mammalian homolog Pes1, is responsible for the interaction with Erb1 or its mammalian homolog Bop1.38, 39 In particular, Erb1 contains multiple phosphoserines based on in‐depth proteomic analysis,40, 41, 42 and the BRCT domain is known to recognize a phosphopeptide.43 Consistent with its scaffolding role suggested by our structural observations, Nop7 assembles into early 90S particles containing 35S pre‐rRNA and likely functions for several steps.34, 38 An additional piece of supporting data is that a temperature sensitive nop7 mutation is synthetically lethal with mutations in Drs1, a DEAD box helicase necessary for the pre‐60S assembly.44, 45 Therefore, similar to Nop53, Nop7 might also be involved in recruiting additional factors required for the function of the pre‐rRNA processing machinery.

In summary, we used a hybrid approach by combining structural information from multiple sources, and successfully built atomic models for five assembly factors in the ITS2 region of the pre‐60S ribosome. The structures provide essential information for the design of future biochemical and genetic experiments to study the mechanisms of eukaryotic ribosome biogenesis.

Materials and Methods

The purification of the pre‐60S Nog2‐particle, chemical cross‐linking of proteins coupled with mass spectrometry analysis (CXMS), and cryo‐EM 3D reconstruction were described previously.26

Modeling of ITS2 factors

Sequences of Cic1, Nop7, Nop15, Nop53, and Rlp7 were subjected to secondary and tertiary structural prediction, using PSIPRED30 and I‐TASSER,31 respectively. For each factor, their core domains were used for rigid‐body fitting in Chimera,46 followed by model rebuilding manually in Coot.35 Information from secondary structural predication was used to aid main‐chain tracing. In certain regions, poly‐alanine models were built first, and sequence assignments were aided by well‐resolved bulky residues such as Phe, Tyr, Trp, and Arg. As described in the main text, factors with known domain structures, such as Cic1, Nop15, and Rlp7 were built first. The location of additional factors in the region of their identified neighbors was facilitated by CXMS data. The CXMS data used for model building are summarized in Supporting Information Table S1. The final coordinates for the five assembly factors are included in the protein databank entry 3JCT.

Supporting information

Supporting Information Table 1.